Research map / Optimization / Optimizers
A Mathematical Model for Curriculum Learning for Parities
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a two-step curriculum, beginning with a biased distribution and switching to the uniform distribution | 2-layer ReLU network learning a 20-parity over 100 bits · 20-parity over 100 bits | convergence time: 16,000 epochs; no curriculum does not achieve non-trivial correlation in 25,000 epochs | Helped |
| Use a continuous curriculum instead of a sharp two-step switch | 2-layer ReLU network learning a 20-parity over 100 bits · 20-parity over 100 bits | test error under the uniform distribution | Mixed |
| Vary the initial bias in a two-step curriculum (ablation) | 2-layer ReLU network learning a 10-parity over 100 bits · 10-parity over 100 bits | convergence time: p1 = 0.001, 0.05,0.1,0.15,...,0.95,0.999; 100,000 | Mixed |
| Use a bounded-step curriculum for Hamming mixtures | Noisy gradient descent on fully connected networks · Hamming mixtures | correlation with target under the uniform distribution: o_d(1) | Hurt |
| Use a two-step curriculum with an initial bias near 1/2 (ablation) | 2-layer ReLU network learning a 10-parity over 100 bits · 10-parity over 100 bits | convergence time: 100,000 | Hurt |
Closest papers by meaning
- Learning to Solve Complex Problems via Dataset Decomposition (NeurIPS 2025 · similarity 0.86)
- Progressive distillation induces an implicit curriculum (ICLR 2025 · similarity 0.85)
- Dual-Level Curriculum Meta-Learning for Noisy Few-Shot Learning Tasks (AAAI 2024 · similarity 0.84)
- CurBench: Curriculum Learning Benchmark (ICML 2024 · similarity 0.84)
- CurES: From Gradient Analysis to Efficient Curriculum Learning for Reasoning LLMs (ICLR 2026 · similarity 0.83)
- Montessori-Instruct: Generate Influential Training Data Tailored for Student Learning (ICLR 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.